Back to skill

Security audit

商业自动中枢

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent business-automation guide, but it can steer an agent toward command execution, scheduled jobs, external integrations, and automatic financial approvals without strong default confirmation controls.

Install only if you want an agent to help design and potentially implement business automations. Keep it in planning or dry-run mode by default, require explicit approval before exec commands, cron creation, finance or CRM writes, notifications, content publishing, or payment steps, and add human approval for new vendors, low-confidence extraction, missing PO data, and all production payment actions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Vague Triggers

Medium
Confidence
91% confidence
Finding
The quick-start triggers are broad natural-language phrases such as '帮我把[系统A]和[系统B]集成起来' and '建一个cron任务' that can match ordinary user requests and cause this high-impact skill to activate unexpectedly. Because the skill advertises exec, scheduling, integration, alerting, and file-writing behavior, accidental invocation can expand the agent's operational scope without an explicit user acknowledgement.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill declares tools including exec and later documents command execution, scheduling, notifications, and writing results to logs, but the description area does not prominently warn users that using the skill can lead to system-impacting actions. This increases the chance that users invoke the skill for planning purposes while the agent interprets the request as authorization to execute commands, create cron jobs, or write files.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
type: condition
      rules:
        - condition: "amount <= 5000"
          goto: auto_approve
        - condition: "default"
          goto: manager_review
  error_handling:
Confidence
87% confidence
Finding
auto_approve

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.